Models Start Late

By the time an architect joins some projects, feasibility work has already set assumptions about site capacity, unit mix, parking and budget. That sequence can make design feel like execution, even though many assumptions are still uncertain.

Models start late

Assumptions, not data. The pro forma targets 120 units, 96,000 net rentable ft², 82% efficiency and a 6.5% yield on cost ($36M cost, $2.34M income, 5.5% cap rate). The design starts short by the starting gap on units and area, 0.3× that on efficiency and 1.5× that on yield. The first change closes the chosen share of the gap; each later change closes 0.8× the share before. Drawings issue at iteration 8; one change is made per iteration after the model starts. A change before issue costs $60k in fees and delay; the first after issue costs 3× that, and each iteration after multiplies the cost by 1.25.

Those early decisions often come from market studies, lender requirements and contractor estimates. They are necessary to decide whether to pursue a site, but they are not the same as a developed design. Testing them against a real plan can reveal conflicts and opportunities that a spreadsheet alone misses.

The pro forma sets KPIs before anyone draws, and the architect is held to them. The study assumes targets of 120 units, 96,000 net rentable ft², 82% efficiency and a 6.5% yield on cost, and a first design that starts 20% short. Model from iteration 0 and 4 changes take the gap from 20 to 3.9 points for $240k. That leaves 115 units, 92,297 ft² and a 6.12% yield on cost.

Each change closes a share of what is left, and the share shrinks. Here the first closes 45% of the gap, the next 36%, the next 29%. Assume drawings issue at iteration 8. If the model starts at iteration 2, 6 changes all land before drawings issue and leave 2.7 points for $360k. The 7th would recover $202k of value but costs $180k. The 8th recovers $142k and costs $225k.

Start two iterations later, at iteration 4, and the same 6 changes recover the same 2.7 points. Two of them now land after drawings are issued, so the bill is $645k instead of $360k. Change 7 would be the first to cost more than it recovers.

At iteration 6 only two changes fit before issue. Making 8 changes still takes the gap to 2.1 points, but they cost $2.15M. Change 6 is the first that costs more than it recovers: $352k against $296k.

At iteration 7 one change fits before issue. Making 10 changes reaches 1.9 points for $4.7M. Change 6 is again the first to lose money, at $439k against $296k, and every change after it recovers less and costs more.

When the model starts as drawings issue, at iteration 8, every change reworks issued sheets. 12 changes cost $9.76M, and the gap is still 1.7 points: 118 units, 94,365 ft² and a 6.33% yield on cost. No number of changes closes it to zero, because each one closes a smaller share of a smaller gap. The architect is still chasing KPIs that were set before the model existed.

The point is not that architecture should happen before every financial decision. It is that design has useful work to do while feasibility is being shaped. Earlier collaboration can make the project’s assumptions more spatially grounded and expose which choices remain open. A schematic plan can test those assumptions before they harden into a brief that is difficult to revisit.